In this paper, we study an important paradox of Deaton and Paxson (1998): Using the Barten model of economies of scale, Deaton and Paxson predict that the share of food consumption in total expenditure should increase as household size rises with per capita expenditure held constant, while their empirical evidence shows the opposite. This paradox has important implications for the measurement of poverty and inequality. Consistent with predictions of the Barten model, we show that the food share in food and a more public good (housing) increases with family size, and that the elasticity of the share of food with respect to family size is larger for poorer households. Therefore, the Deaton-Paxson Paradox cannot simply be due to an incorrect theoretical model. We provide evidence that food preparation time may be important in understanding this puzzle. Key Words: household consumption, economies of scale JEL Classification: D1, E2 # The corresponding author. Email: [email protected]; phone: (512)475-8540. + We thank participants of the Lonestar Conference, Peter Debaere, Don Fullerton, Dan Hamermesh, Gerald Oettinger, Dan Slesnick, Max Stinchcombe and Clayton Vernon for their helpful comments. We especially would like to thank Angus Deaton for his comments on an earlier version of this paper. All remaining errors are ours. 1